To be clear, the property we want is, broadly speaking, "hard to tamper with (at scale) without getting caught". Ideally we would also catch tampering at small scale, but that's darn close to incompatible with denying vote buying. (A possible resolution is, of course, "well, what's wrong with vote buying between informed consenting adults?" But the unfortunate history shows that our electoral systems must compensate for underinformed coerced voters, and their regulations are written in blood.)
Existing voting systems do have a countermeasure, if not a fantastic one: creating a trail of physical artifacts that can be manually audited to verify vote totals, and a roll of accepted votes to compare the count of said physical artifacts against. It's not fantastic because the error rates on those physical artifacts are stupendous, but tampering with votes at scale can then require (a) physical access, which humans are well-equipped to reason about, and (b) generating and destroying big piles of said physical artifact, which is expensive and expensive to hide. The gold standard of tampering with physical elections that we know of is basically denying observers the chance to audit, which is rightly considered suspicious.
Tampering with electronic votes at scale does not have these cost properties. We can magnify costs without giving voters the ability to prove their votes to a third party (I am aware there are probabilistic constructions), but all such constructions (a) are much harder for the average voter to reason about than monitoring physical access (humans are quite optimized for monitoring physical access), yet (b) requires voters to actually audit their own votes and report non-inclusion en masse. You can see why this is a non-starter.